🐻⬇️🏀

2025-2026 NBA Model Performance Analysis

Scope
All Games Ranked Games

All scored games in the selected league and season. AP Poll is excluded here.

Season
2025-2026

Comparing prediction accuracy across 1305 games using multiple rating models.

Model Catalog

7-day holdout coverage: 16/17 models .

Rolling Holdout Curves

Each point is a strict weekly holdout: train on all games before that week, test on that week. This first version uses a 21-day warmup, then 7-day holdouts stepped forward weekly.

Log Loss Brier AUC Accuracy

Weekly strict holdout log loss. Lower is better. Showing 16 models across 33 windows. Click legend items to hide/show series.

Recent Window Winners

Holdout Best Log Loss Runner-up Models
Jun 3 Elo 0.636 Bradley-Terry Recency (0.687) 16
May 27 - Jun 2 Home Team Baseline 0.714 Points Off/Def Recency (0.718) 16
May 20 - May 26 Adjusted Context Blend 0.551 Log Adjusted (0.551) 16
May 13 - May 19 Elo 0.615 Bradley-Terry Recency (0.655) 16
May 6 - May 12 Log Adjusted 0.495 Adjusted Efficiency (0.496) 16
Apr 29 - May 5 Points Off/Def 0.654 Margin (0.654) 16
Apr 22 - Apr 28 Adjusted Context Blend 0.589 Bradley-Terry (0.590) 16
Apr 15 - Apr 21 Home Team Baseline 0.646 Bradley-Terry Recency (0.709) 16

Model Performance Leaderboard

Models ranked by strict holdout AUC when available (fallback: full-season AUC). Hover over column headers for explanations.

# Model 7d Split AUC Acc Brier LogLoss n AUC 7d Acc 7d Brier 7d n 7d
1 Avg Margin Baseline Avg Margin Baseline Predict from simple average scoring margin in the training window. More → STRICT
3g
 0.757 69.1% 0.203 0.593 1173 - 0.0% 0.287 3
2 Efficiency Efficiency Tempo-adjusted efficiency version of Pythagorean ratings. More → FULL
no 7d
 0.750 67.8% 0.204 0.593 1173 - - - 0
3 Points Off/Def Points Off/Def Raw points regression with separate offensive and defensive team parameters. More → STRICT
3g
 0.749 67.3% 0.209 0.607 1173 - 0.0% 0.304 3
4 Pythagorean Pythagorean Pythagorean win expectation from raw points scored and allowed. More → STRICT
3g
 0.748 68.2% 0.215 0.622 1173 - 0.0% 0.262 3
5 Adjusted Efficiency Adjusted Efficiency Opponent-adjusted efficiency model with separate offensive and defensive components. More → STRICT
3g
 0.747 67.8% 0.205 0.596 1173 - 0.0% 0.308 3
6 Log Adjusted Log Adjusted Log-scale adjusted efficiency model that downweights blowout leverage. More → STRICT
3g
 0.747 67.8% 0.205 0.596 1173 - 0.0% 0.308 3
7 Bradley-Terry Bradley-Terry Static logistic paired-comparison model with one team strength parameter. More → STRICT
3g
 0.746 67.9% 0.206 0.601 1173 - 0.0% 0.332 3
8 Margin Margin Linear team-strength model fit on point differential instead of binary wins. More → STRICT
3g
 0.740 67.4% 0.212 0.613 1173 - 0.0% 0.305 3
9 Elo Elo Streaming paired-comparison rating with recency baked into sequential updates. More → STRICT
3g
 0.738 68.1% 0.208 0.604 1173 - 33.3% 0.293 3
10 Home Team Baseline Home Team Baseline Always favor the home team with a fixed prior. More → STRICT
3g
 0.550 55.0% 0.250 0.693 1173 - 33.3% 0.293 3
- Bradley-Terry Recency Bradley-Terry Recency Static Bradley-Terry with exponential recency weights on newer games. More → STRICT
3g
 - - - - 0 - 33.3% 0.299 3
- Margin Recency Margin Recency Margin regression with exponential recency weights on newer games. More → STRICT
3g
 - - - - 0 - 66.7% 0.251 3
- Points Off/Def Recency Points Off/Def Recency Off/def points regression with exponential recency weights. More → STRICT
3g
 - - - - 0 - 66.7% 0.250 3
- Core Ensemble Core Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and points off/def. More → STRICT
3g
 - - - - 0 - 33.3% 0.288 3
- Recency Ensemble Recency Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and recency points off/def. More → STRICT
3g
 - - - - 0 - 33.3% 0.278 3
- Dynamic Bradley-Terry Dynamic Bradley-Terry Time-evolving paired-comparison model with latent team strength drift. More → STRICT
3g
 - - - - 0 - 0.0% 0.291 3
- Adjusted Context Blend Adjusted Context Blend Experimental context-heavy win model blending strong team components with rest and venue context. More → STRICT
3g
 - - - - 0 - 0.0% 0.352 3

Methodology

ELO / Bradley-Terry

  • ELO: Iterative updates, K=64, HCA=100
  • BT: Static logistic regression on all games
  • Both model win probability, not margin
  • ELO updates after each game; BT fits once

Pythagorean Models

  • Raw: Classic points scored/allowed formula
  • Efficiency: Pace-adjusted (pts per possession)
  • Adjusted: Opponent-adjusted efficiency
  • Log: Log-linear multiplicative scale

Margin Regression

  • Team-level ridge regression on point margin
  • Linear Bradley-Terry (margin target)
  • Alpha=0.05 (CV-tuned)
  • Learns home advantage from data (~6 pts)

Baselines

  • Home Team: Always predict home wins (60%)
  • Avg Margin: Higher average margin wins
  • Models should beat these to add value